Engenharia de tráfego em domínio MPLS utilizando técnicas de inteligência computacional

Abstract

The current need of simultaneous transmission of data, voice, and video has changed the target of telecommunication network technologies. Instead of providing only one type of service, networks must offer different levels of Quality of Service (QoS), demanding new techniques for its management. One of these techniques is Traffic Engineering, which dynamically manages the traffic distribution in the network, minimizing congestions, instabilities or loss of the agreed QoS. On the other hand, the growing complexity of present-day networks turns the human intervention into a weak point of the process of management and control. It requires that networks be able to perform some kind of intelligent behavior, exhibiting characteristics like adaptability, fault tolerance and robustness to environment variations. This work proposes and develops a Traffic Engineering system, capable of supporting mixed traffic (data, voice, and video) with different levels of QoS in the network, using MPLS, principles of Autonomic Computing and techniques of Computational Intelligence. While MPLS permits an effective implementation of traffic control and QoS, Autonomic Computing allows the network to automatically react to environment changes during its operation, producing a self-management behavior. In complement, monitoring data are processed using Computational Intelligence heuristics, like Fuzzy Logic, Artificial Neural Networks, and Genetic Algorithms, which help in taking the necessary decisions for reactive and proactive actions. The system was tested using the network simulator ns2, with different scenarios of data, voice, and video traffic crossing a MPLS domain. The results have shown that the system is capable of implementing new routes, considering the application QoS requirements and the available network resources, and blocking new admissions that could not be attended. Adaptation has also been observed, as replacement routes were provided in case of decay or loss of present routes, change in application requirements, or availability of better routes. In comparison to situations without Traffic Engineering, the implemented system has shown improvements of 44% in the average network utilization and 52% in the offered and delivered load ratio, for the worst cases. The average packet loss was reduced to near zero.